This repository provides a reinforcement learning environment for high-frequency trading, specifically designed for developing interest rate trading strategies using historical order book data. It targets quantitative finance researchers and developers interested in algorithmic trading without prior market dynamic assumptions. The framework allows agents to learn directly from data, offering a simulator for agent interaction and experience gain.

How It Works

The project utilizes a reinforcement learning agent interacting with a custom-built market simulator. This simulator processes historical high-frequency order book data, allowing the agent to learn trading strategies without explicit market dynamic modeling. The architecture is inspired by Udacity's Smartcab and OpenAI Gym, providing a familiar structure for RL practitioners.

Quick Start & Requirements

• Install using Python 2.7.
• Required libraries: Bintrees, Matplotlib, NumPy, Pandas, Seaborn, BeautifulSoup.
• Run command: python -m market_sim.agent [-h] [-t] [-d] [-s] [-m] <OPTION>
• Data: Unzip example datasets into data/preprocessed/.
• Note: Simulations may take several minutes.

Highlighted Details

• Framework developed for a master's thesis in quantitative finance.
• Focuses on interest rate trading using order book data.
• Employs tile coding for function approximation in RL.

Maintenance & Community

• Primarily a personal research project; no explicit community channels or active maintenance signals are present.

Licensing & Compatibility

• License: Apache 2.0.
• Compatible with commercial use and closed-source linking.

Limitations & Caveats

The project requires Python 2.7, which is end-of-life and may present compatibility issues with modern systems and libraries. The README indicates simulations can take "several minutes," suggesting potential performance bottlenecks for extensive testing.